Page 1 of 1

NEB Convergence problem

Posted: Wed Jul 27, 2022 12:31 pm
by Hongbo
Dear Professor Graeme Henkelman,

I'm trying to do a cNEB calculation of CO oxidation on In2O3. I use IOPT = 1, EDIFF = 1E-7, EDIFFG = -5E-02. The calculation didn't converge. I changed the paremeters to IOPT = 2 or 3, EDIFF = 1E-5, EDIFFG = -1E-01. Still it can't converge. I use images = 5. In another calculation In use mages = 10. I failed to converge,either. I attached my complete calculation files here. Could you please give me some comment about it?

Sincerely,

Hongbo

Re: NEB Convergence problem

Posted: Wed Jul 27, 2022 1:41 pm
by graeme
This calculation look reasonable. The only issue I see is very high forces at the start of the calculation. Switch to a TIMESTEP of 0.01 and run about 20 iterations and make sure that the forces systematically drop from ~30 eV/Ang to <1 eV/Ang. Once all the forces are below 1 eV/Ang, you can go back to a TIMESTEP of 0.1. You could also use the IDPP method to set up your initial band with lower initial forces. This is easy to do and costs nothing; you can use the method within ASE or our nebmake.py script which will automatically use the IDPP method.

Re: NEB Convergence problem

Posted: Sun Jul 31, 2022 12:24 am
by Hongbo
Thank you Professor Graeme Henkelman. The problem solved by using a TIMESTEP of 0.01 first, and later using TIMESTEP of 0.05. But if I use a TIMESTEP of 0.01 first, and later I use TIMESTEP of 0.1, still it can't converged.

Re: NEB Convergence problem

Posted: Sun Jul 31, 2022 12:25 am
by graeme
Ok, it sounds like you have what you need to make things work. Great!